Geometry of Waves

Waveguides Derived
Wave-equation:
#2 y #x 2

(Smith, Cook, et al, 83+)

" c12

#2 y #t 2

!0

General form of solution (dAlembert): Georg Essl

georg@mle.media.mit.edu

y ( x, t ) ! f1 ( x $ ct ) $ f 2 ( x " ct )

Standing and Traveling Waves

Standing Waves via sums of traveling waves

Standing and Traveling Waves

Standing Waves via sums of traveling waves

Bessel Function
J0(k|r|) = 0 are frequencies
1
1

Bessel Function
J0(k|r|) = 0 are frequencies
1
1

0.5

0.5

0.5

0.5

0.5

0.5

8 x

10

12

14

16

0.1

0.2

0.3

0.4

0.5 x

0.6

0.7

0.8

0.9

0.5

0.5

8 x

10

12

14

16

0.1

0.2

0.3

0.4

0.5 x

0.6

0.7

0.8

0.9

Where is my analytic formula?

Where is my analytic formula? !

After 120 years of acoustics, the zeros or frequencies are still not understood. (Zelditch03)

(Keller & Rubinov60)

Isolated Membrane Oscillations

How Big A Mistake?

Discovered that the error even for low frequencies is small m 0 n 1 2 3 4 1 2 3 4 error 2.0% 0.4% 0.2% 0.1% 1.0% 0.3% 0.1% 0.1%

(0,1)

(0,2)

(0,3)

1
(1,1) (1,2) (2,1)

(mth Bessel, nth zero)

Schlieren Visualization of Liquid Tank

Topology of a Square Drum

Glue all reflecting edges together Discontinuous lines with difficult connectivity bookkeeping Straight lines and immediate connectivity

(Chinnery, Humphrey and Beckett97)

Homeomorphism: Cartesian to Cylindrical Coordinates

(Keller & Rubinov60)

Paths in a Circular Room

Resonance is a path in a room that returns to its starting point

Resonant Paths
2:1=1 3:1=1.5662 4:2=1.8735 4:1=2.2508 5:2=2.3891

Unreachable Spots in Circular Rooms

2:1=1 3:1=1.5662 4:2=1.8735 4:1=2.2508 5:2=2.3891

6:3=2.7471

5:1=2.9741

6:2=3.0059

7:3=3.2403

8:4=3.6206

6:3=2.7471

5:1=2.9741

6:2=3.0059

7:3=3.2403

8:4=3.6206

7:2=3.6757

6:1=3.7132

8:3=3.8167

9:4=4.1014

8:2=4.3752

7:2=3.6757

6:1=3.7132

8:3=3.8167

9:4=4.1014

8:2=4.3752

9:3=4.4457

7:1=4.4601

10:5=4.4941

10:4=4.6517

9:2=5.0927

9:3=4.4457

7:1=4.4601

10:5=4.4941

10:4=4.6517

9:2=5.0927

10:3=5.1092

8:1=5.2111

10:2=5.8217

9:1=5.9645

10:1=6.7194

10:3=5.1092

8:1=5.2111

10:2=5.8217

9:1=5.9645

10:1=6.7194

(Gbur et al. 02)

Gouy Phase
Phase shift through focal point But not for wavelengths longer than focal length!

What a catastrophe: Caustic

Morning Coffee Caustic Cardioid-like shape Focal regions of reflected wave rays
(Bergervoet) (Picture courtesy of Erin Panttaja)

Along the axis of symmetry, at points where the intensity is zero, the phase is discontinuous.

Short wavelengths, its pi/2. Very large wavelengths it vanishes. Nobody understands the in-between.

What a catastrophe: Caustic

Morning Coffee Caustic Cardioid-like shape Focal regions of reflected wave rays
(Picture courtesy of Erin Panttaja)

Construction
Domain

Construction
Domain Interaction Point

Construction
Domain Interaction Point Pencil of Rays

Construction
Domain Interaction Point Pencil of Rays Ray Angles

Construction
Domain Interaction Point Pencil of Rays Ray Angles Segment Distance

Holditchs Caustic (First Order)

Cusp (Semi-cubical)

Lifted Cusp

Huygens & Malus

Final front as sum of circular expansions Front normal to ray Maintain normal through reflections

Construction
Domain Interaction Point Pencil of Rays Ray Angles Segment Distance Front Set

Construction
Domain Interaction Point Pencil of Rays Ray Angles Segment Distance Front Set

Waves and Rays

Pearcey Diffraction

(Berry)

Lineland

Traveling on the Line is a Loop

Lift double line to loop

The Topology

Lift reflection states to double-loop

Double Loops indicate Dynamical Properties

Interactions in Lineland

Dirichlet

Neumann

Mixed

Interaction Modeling

Integration properties Frequency properties

A Notion of Coincidence

Excitation Points in 1-D

Center Excitation

Far Off-Center Excitation

1-D Topology Examples

Strings & Tubes Marimba Bars
" Same topology " Different propagation

Flatland
Some Membranes and Plates are

Wine Glasses Tibetan Singing Bowls

" Similar topology " Same dynamics

Nice Rooms in 2-D

(Abbott1884)

Circular cylinder = string

(but less nice than in 1-D)

2-D Outline
Background
" Numerical Efficiency
Why we want to walk on lines

Flatland

The topology
" How we walk on lines in 2-D " On Drums, round and square
Why 3-D can be nicer than 2-D Why 2-D is tricky

Background

" How we interact in 2-D

Rooms in Flatland can be expensive

Can we have a Flat room thats cheap?

Mesh Methods: O(n2)

The cost of walking on a line is independent of its size! Are there ways to walk a flat room via lines?

Rays and Wavefronts

Short Wavelengths localize well Ray Assumption Short-wave Asymptotics Hamiltonian Room Well run with it and see

Flatland

Waterfronts

Ray

The Topology

Rectangular Flatland

Covering Space of a Square Drum

Single Ray Ray Family

On Square Drums

Path-connected orbit on a Torus

Path-connected orbit on a Torus

Modes and Orbits on Tori

Winding around torus for closed orbits: length of orbit Index counting the number of special points (boundary reflections): phase changes there. -> Resonance (Chazarain74, DuistermaatGuillemin75)

Circular Flatland

On Circular Drums

Drum!
Circular Membrane Uniform (or nonuniform) thickness

Interacting in Flatland

Interaction Modeling

Similar paths on torus

Center Excitation

Torus does not itself provide reference Mirror symmetric Rotationally symmetric

Center Excitation

Off-Center Excitation (1)

Off-Center Excitation (1)

Off-Center Excitation (2)

Off-Center Excitation (2)

Stability of Coincidence Knots under Perturbation

Stable!

Unstable!

Demo

Open Problems
Relationship of topological spectrum and conventional spectrum What is the proper treatment of excitations on caustic? (unknown) Can asymptotic assumption be removed? (hard) Domain shape (guitar, violin top plates, ) (Kac 66: Can we hear the shape of a drum? hard!)

Excitation on different points of the Torus

The end (for now)

Nice Room in Flatland

At last, to complete a series of minor indiscretions, at a meeting of our Local Speculative Society held at the palace of the Prefect himself, - some extremely silly person having read an elaborate paper exhibiting the precise reasons why Providence has limited the number of Dimensions to Two.

Questions?

- Edwin A. Abbott, Flatland: A Romance of Many Dimensions

Burning Regions in a Circular Room

Caustics is formed by ray family reflecting at one angle

If you know the Resonance, correct your stride

(Essl&Cook02-04)

You may have to skip a step to get the Resonance

Why bother?
Actual science to be had Faster and more furious Talks with pretty pictures (And cool new math)
(Robinson)

Vibration of a Rectangular Drum

Frequencies function of the squareroot of dimensions Shape sinusoidal

(2,1)

(1,1)

(2,2) (1,2)

Chladni Figures - Square Drum

Sand on membrane and proper finger placement -> Chladni figures

What to do?
We lie, cheat, steal.

(Chladni, Heller)

What to do?
We lie, cheat, steal. Triple-A
" Adjust " Approximate " Asymptotics
Expansions of Bessel function known Expansions of Short-Wave Asymptotics known Same!

High Mode Troubles

High modes dense High modes distribution look random Berry-Tabor conjecture: Use Poisson tail! (Like Schroeder does for roomacoustics) Alternative: Low-resolution mesh (Banded Waveguide Mesh, Serafin & Smith 01)

Interaction Points (Covering Space)

Need reference to boundary (how close to reflection) Load at four mirror symmetric points (string we had two)

Toroidal Topologies
Introduction: Rectangular Membrane
" Ray methods & Topology

Challenge: Circular Membranes

" Bessel, Asymptotics and other catastrophes

Application: Interaction Modeling

Nice Rooms in Flatland

I call our world Flatland, not because we call it so, but to make its nature clearer to you, my happy readers, who are privileged to live in Space.
- Edwin A. Abbott, Flatland: A Romance of Many Dimensions, 1884

A Rough Route Through Flatland

Whats so cool about studying flat rooms? Nice Rooms in 1-D
" The lines and circles of strings, tubes, bars, shells and bells

Nice Rooms in 2-D

" Square, round and on doughnuts (2-D)

What makes a nice room?

Symmetric Simple Nice sounding Efficiently computable

Whats so cool about studying flat rooms?

Interactive Applications:
" Music Performance " Games " Motion Pictures (Especially CG)

Requirements:
" " " " Interactive Rates (44100 iterations/second) Commodity Hardware Proper Numerical Behavior Ease of Use for Nonexperts

Publications
Banded Waveguides
" Solid Bars: Essl & Cook99, Essl & Cool00 " Glasses & Bowls: Essl & Cool02, Serafin et al02 " Membranes: Essl02, Kapur et al02, Essl et al04 " Interactions: Essl04 submitted

Circular Rooms in Flatland

But Its harder to live in a round room in flatland

(Veselov02)

Huygens in Flatland

Whats so hard about it?

Huygens Principle

Initial disturbance Final front as sum of circular expansions

" Proven to not hold in 1-D, 2-D (us!) and 2*n-D (all even dimensional rooms)

Only odd dimensional rooms of 3 or more are generically nice!

(Watson22)

Whats so hard about it?

Restful places in a round room are the zeros of the Bessel function
" Special means rich and difficult in certain situations, unfortunately this is one !

Many paths sound the same

2:1=1 3:1=1.5662 4:2=1.8735 4:1=2.2508 5:2=2.3891

6:3=2.7471

5:1=2.9741

6:2=3.0059

7:3=3.2403

8:4=3.6206

7:2=3.6757

6:1=3.7132

8:3=3.8167

9:4=4.1014

8:2=4.3752

9:3=4.4457

7:1=4.4601

10:5=4.4941

10:4=4.6517

9:2=5.0927

10:3=5.1092

8:1=5.2111

10:2=5.8217

9:1=5.9645

10:1=6.7194

Covering of Ray Families

2:1=1 3:1=1.5662 4:2=1.8735 4:1=2.2508 5:2=2.3891

The end (for now)

6:3=2.7471

5:1=2.9741

6:2=3.0059

7:3=3.2403

8:4=3.6206

7:2=3.6757

6:1=3.7132

8:3=3.8167

9:4=4.1014

8:2=4.3752

9:3=4.4457

7:1=4.4601

10:5=4.4941

10:4=4.6517

9:2=5.0927

10:3=5.1092

8:1=5.2111

10:2=5.8217

9:1=5.9645

10:1=6.7194